Bug 218886 - ifconfig lagg0 destroy will crash the system
Summary: ifconfig lagg0 destroy will crash the system
Status: New
Alias: None
Product: Base System
Classification: Unclassified
Component: kern (show other bugs)
Version: 11.0-RELEASE
Hardware: amd64 Any
: --- Affects Some People
Assignee: freebsd-net (Nobody)
Depends on:
Reported: 2017-04-26 13:10 UTC by Lutz Rabing
Modified: 2017-05-26 13:06 UTC (History)
3 users (show)

See Also:

crash_screen (643.08 KB, image/png)
2017-05-02 13:08 UTC, Lutz Rabing
no flags Details
core.txt.0 (116.33 KB, text/plain)
2017-05-02 15:08 UTC, Lutz Rabing
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Lutz Rabing 2017-04-26 13:10:11 UTC
When I destroy a lagg interface consisting of ix0 and ix1 the system will
crash immmediatetly.

command used:  ifconfig lagg0 destroy

System: FreeBSD 11.0-RELEASE-p9 (amd64)
Comment 1 Alan Somers freebsd_committer 2017-04-26 17:02:36 UTC
Lutz, please post a stack trace.  You can find it in /var/crash/core.txt.  It will also be printed to the screen when the crash happens.
Comment 2 Lutz Rabing 2017-05-02 13:08:05 UTC
Created attachment 182244 [details]
Comment 3 Lutz Rabing 2017-05-02 13:09:20 UTC
Hi Alan,

the /var/crash/ dir is empty. The system appears to be dead immediately.
I just have IPMI console access and the error message appears for 0.5s
before the BIOS boot screen is shown. I'll attach a screenshot.

Here is the config of the lagg interface:

lagg0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
        ether 22:22:22:22:20:05
        inet6 fe80::92e2:baff:fe61:c14c%lagg1 prefixlen 64 scopeid 0x11
        inet netmask 0xffffff00 broadcast
        nd6 options=21<PERFORMNUD,AUTO_LINKLOCAL>
        media: Ethernet autoselect
        status: no carrier
        groups: lagg
        laggproto lacp lagghash l2,l3,l4
        laggport: ix2 flags=0<>
        laggport: ix3 flags=0<>
Comment 4 Alan Somers freebsd_committer 2017-05-02 14:22:02 UTC
Lutz, unfortunately that screenshot doesn't capture anything useful.  It seems like your system is not setup to dump core on panic.  Could you please check the output of "dumpon -l"?  If it's empty, then you don't have a dump device set.  Normally, you can put "dumpdev=AUTO" in /etc/rc.conf, which will use your first swap partition as a dump device.
Comment 5 Lutz Rabing 2017-05-02 14:39:59 UTC
(In reply to Alan Somers from comment #4)
Yes, dumpdev="NO"  was set in rc.conf. I changed it to AUTO and
repeated the crash. There is just a info.0 and vmcore.0:

drwxr-x---   2 root  wheel           8 May  2 16:34 .
drwxr-xr-x  25 root  wheel          25 May  2 18:34 ..
-rw-r--r--   1 root  wheel           2 May  2 16:34 bounds
-rw-------   1 root  wheel         452 May  2 16:34 info.0
lrwxr-xr-x   1 root  wheel           6 May  2 16:34 info.last -> info.0
-rw-r--r--   1 root  wheel           5 Nov 11  2014 minfree
-rw-------   1 root  wheel  1013112832 May  2 16:34 vmcore.0
lrwxr-xr-x   1 root  wheel           8 May  2 16:34 vmcore.last -> vmcore.0

# cat info.0
Dump header from device: /dev/gpt/swap2
  Architecture: amd64
  Architecture Version: 2
  Dump Length: 1013112832
  Blocksize: 512
  Dumptime: Tue May  2 16:33:14 2017
  Hostname: core10.omc.net
  Magic: FreeBSD Kernel Dump
  Version String: FreeBSD 11.0-RELEASE-p9 #0: Tue Apr 11 08:48:40 UTC 2017
  Panic String: page fault
  Dump Parity: 3897632610
  Bounds: 0
  Dump Status: good
Comment 6 Alan Somers freebsd_committer 2017-05-02 14:43:40 UTC
Progress!  Now try running "crashinfo".  Normally that should run automatically, unless you put "crashinfo_enable=NO" in /etc/rc.conf.  It could also fail to run if your kernel is stored in a nondefault location.
Comment 7 Lutz Rabing 2017-05-02 15:08:49 UTC
Created attachment 182246 [details]
Comment 8 Alan Somers freebsd_committer 2017-05-02 15:26:55 UTC
It looks like this was fixed in head by svn r312979 by loos, but that change did not make it into 11.0-RELEASE.  You could try installing a stable/11 kernel.  Alternatively, perhaps loos, CCed could suggest a workaround.